select top 1 * from c returns count size as num of partitions * actual query count in cosmosdb spark

I am using azure-cosmosdb-spark library for scala and I was trying to query the following

select top 1 * from c

but I gotten the final dataframe count to be 8 instead of 1. I suspect that the CosmosDBRDDIterator splits it into multiple partitions (in this case 8) and processed the query.

The result count will always be 8x the count of the actual results regardless the select query executed.

Is there any way I can avoid this and get my actual count as 1?